Amir Abbasnejad1,2, Oren Tomkins-Netzer3, Aaron Winter4, Alon Friedman2,5, Alan Cruess4, Yonatan Serlin6, Jaime Levy7. 1. Faculty of Computer Science, Dalhousie University, Halifax, NS, Canada. 2. Emagix, Inc., Halifax, NS, Canada. 3. Department of Ophthalmology, Faculty of Medicine, Carmel Medical Center, Technion, Haifa, Israel. 4. Department of Ophthalmology, QEII Hospital, Dalhousie University, Halifax, NS, Canada. 5. Departments of Medical Neuroscience and Pediatrics, Faculty of Medicine, Dalhousie University, Halifax, NS, Canada. 6. Neurology Residency Training Program and Department of Neurology and Neurosurgery, Jewish General Hospital (J.M.), McGill University, Montreal, QC, Canada. 7. Department of Ophthalmology, Hadassah-Hebrew University Medical Center, Jerusalem, Israel. levjaime@gmail.com.
Abstract
OBJECTIVE: To present a fluorescein angiography (FA)‒based computer algorithm for quantifying retinal blood flow, perfusion, and permeability, in patients with diabetic retinopathy (DR). Secondary objectives were to quantitatively assess treatment efficacy following panretinal photocoagulation (PRP) and define thresholds for pathology based on a new retinovascular function (RVF) score for quantifying disease severity. METHODS: FA images of 65 subjects (58 patients and 7 healthy volunteers) were included. Dye intensity kinetics were derived using pixel-wise linear regression as a measure of retinal blood flow, perfusion, and permeability. Maps corresponding to each measure were then generated for each subject and segmented further using an ETDRS grid. Non-parametric statistical analyses were performed between all ETDRS subfields. For 16 patients, the effect of PRP was measured using the same parameters, and an amalgam of RVF was used to create an RVF index. For ten post-treatment patients, the change in FA-derived data was compared to the macular thickness measured using optical coherence tomography. RESULTS: Compared to healthy controls, patients had significantly lower retinal and regional perfusion and flow, as well as higher retinal permeability (p < 0.05). Moreover, retinal flow was inversely correlated with permeability (R = -0.41; p < 0.0001). PRP significantly reduced retinal permeability (p < 0.05). The earliest marker of DR was reduced retinal blood flow, followed by increased permeability. FA-based RVF index was a more sensitive indicator of treatment efficacy than macular thickness. CONCLUSIONS: Our algorithm can be used to quantify retinovascular function, providing an earlier diagnosis and an objective characterisation of disease state, disease progression, and response to treatment.
OBJECTIVE: To present a fluorescein angiography (FA)‒based computer algorithm for quantifying retinal blood flow, perfusion, and permeability, in patients with diabetic retinopathy (DR). Secondary objectives were to quantitatively assess treatment efficacy following panretinal photocoagulation (PRP) and define thresholds for pathology based on a new retinovascular function (RVF) score for quantifying disease severity. METHODS: FA images of 65 subjects (58 patients and 7 healthy volunteers) were included. Dye intensity kinetics were derived using pixel-wise linear regression as a measure of retinal blood flow, perfusion, and permeability. Maps corresponding to each measure were then generated for each subject and segmented further using an ETDRS grid. Non-parametric statistical analyses were performed between all ETDRS subfields. For 16 patients, the effect of PRP was measured using the same parameters, and an amalgam of RVF was used to create an RVF index. For ten post-treatment patients, the change in FA-derived data was compared to the macular thickness measured using optical coherence tomography. RESULTS: Compared to healthy controls, patients had significantly lower retinal and regional perfusion and flow, as well as higher retinal permeability (p < 0.05). Moreover, retinal flow was inversely correlated with permeability (R = -0.41; p < 0.0001). PRP significantly reduced retinal permeability (p < 0.05). The earliest marker of DR was reduced retinal blood flow, followed by increased permeability. FA-based RVF index was a more sensitive indicator of treatment efficacy than macular thickness. CONCLUSIONS: Our algorithm can be used to quantify retinovascular function, providing an earlier diagnosis and an objective characterisation of disease state, disease progression, and response to treatment.